From "Facets of Uncertainty: Epistemic Uncertainty, non-Stationarity, Likelihood, Hypothesis Testing, and Communication," Keith Beven, Hydrological Sciences Journal, Volume 61, No. 9, pp. 1625-1665.
| Type of uncertainty | Description |
| Aleatory | Uncertainty with stationary statistical characteristics. May be structured (bias, autocorrelation, long-term persistence) but can be reduced to a stationary random distribution. |
| Epistemic (System Dynamics) | Uncertainty arising from a lack of knowledge about how to represent the catchment system in terms of both model structure and parameters. Note that this may include things that are included in the perceptual model of the catchment processes but are not included in the model. They may also include things that have not yet been perceived as being important but which might result in reduced model performance when surprise events occur. |
| Epistemic (Forcing and Response Data) | Uncertainty arising from lack of knowledge about the forcing data or the response data with which model outputs can be evaluated. This may be because of commensurability or interpolation issues when not enough information is provided by the observational techniques to adequately describe variables required in the modeling process. May be a function of a limited gauging network, lack of knowledge about how to interpret radar data or non-stationarity and extrapolation in rating curves |
| Epistemic (Disinformation) | Uncertainties in either system representation or forcing data that are known to be inconsistent or wrong. Real surprises. Will have the expectation of introducing disinformation into the modeling processes resulting in biased or incorrect inference (including false positives and false negatives in testing models as hypotheses). |
| Semantic / Linguistic | Uncertainty about what statements or quantities in the relevant domain actually mean. (There are many examples in hydrology including storm runoff, baseflow, hydraulic conductivity, stationarity, etc.) This can partly result from commensurability issues that quantities with the same name have different meanings in different contexts or scales. |
| Ontological | Uncertainty associated with different belief systems. A relevant example here might be beliefs about whether the formal probability is an appropriate framework for the representation of beliefs about the nature of model residuals. Different beliefs about the appropriate assumptions could lead to very different uncertainty estimates so that every uncertainty estimate will be conditional on the underlying beliefs and consequent assumptions. |
